Turbopump fault detection algorithm based on protruding frequency components RMS and SVM

A fault detection algorithm based on protruding frequency components RMS (Root Mean Square) and SVM (Support Vector Machine) was proposed for LRE (Liquid Rocket Engine) turbopump real-time fault detection. In the algorithm, contribution coefficient and contribution limit were proposed to find protruding frequency components which have greater contribution to the frequency spectrum changes than other frequency components. The algorithm firstly divides the signal frequency spectrum into several bands. Secondly, in every frequency band the algorithm selects all the protruding frequency components and computes their RMS value. Thirdly the algorithm combines the RMS values of all frequency bands to construct a multi-dimensional vector as a fault feature. At last the algorithm uses the fault feature vectors got from historical signals to construct the SVM training sample set, and obtains SVM classifier which can be used in turbopump real-time fault detection. We used some historical vibration acceleration signals of a certain type of turbopump to validate the algorithm. The test results showed that the algorithm met the demands of accuracy and real-time ability.